Atomistic Insights into Nucleation and Formation of Hexagonal Boron

Mar 20, 2017 - Tax reform law will boost industry but worries higher ed officials. The most sweeping overhaul of the U.S. tax code in 31 years was sig...
0 downloads 12 Views 10MB Size
Atomistic Insights into Nucleation and Formation of Hexagonal Boron Nitride on Nickel from First-Principles-Based Reactive Molecular Dynamics Simulations Song Liu,† Adri C. T. van Duin,‡ Diana M. van Duin,‡ Bin Liu,*,† and James H. Edgar*,† †

Department of Chemical Engineering, Kansas State University, Durland Hall, Manhattan, Kansas 66506, United States RxFF_Consulting LLC, State College, Pennsylvania 16801, United States



S Supporting Information *

ABSTRACT: Atomistic-scale insights into the growth of a continuous, atomically thin hexagonal boron nitride (hBN) lattice from elemental boron and nitrogen on Ni substrates were obtained from multiscale modeling combining density functional theory (DFT) and reactive molecular dynamics. The quantum mechanical calculations focused on the adsorption and reaction energetics for the hBN building-block species, i.e., atomic B, N, BxNy (x, y = 1, 2), on Ni(111) and Ni(211), and the diffusion pathways of elemental B and N on these slab model surfaces and in the sublayer. B can diffuse competitively on both the surface and in the sublayer, while N diffuses strictly on the substrate surface. The DFT data were then used to generate a classical description of the Ni−B and Ni−N pair interactions within the formulation of the reactive force field, ReaxFF. Using the potential developed from this work, the elementary nucleation and growth process of an hBN monolayer structure from elemental B and N is shown at the atomistic scale. The nucleation initiates from the growth of linear BN chains, which evolve into branched and then hexagonal lattices. Subsequent DFT calculations confirmed the structure evolution energetically and validate the self-consistency of this multiscale modeling framework. On the basis of this framework, the fundamental aspects regarding crystal quality and the role of temperature and substrates used during hBN growth can also be understood. KEYWORDS: 2D materials, hexagonal boron nitride, CVD, DFT, reactive molecular dynamics, ReaxFF force field, growth mechanism

F

thickness, area, crystal perfection, crystallographic orientation, and purity is essential. To prepare atomically thin monolayers, catalytic chemical vapor deposition (CVD) of hBN on transition metal surfaces, such as Ni,13−18 Cu,19−22 Pd,23 and Pt,24 has been employed. However, the presence of high-density grain boundaries and other defects degrades the properties of hBN obtained from CVD.13,18,25 For this reason, a detailed mechanistic understanding of the crystal growth facilitated by substrate surface in a CVD environment is critical to improve the synthesis technique for high-quality hBN materials. There are still unresolved issues regarding hBN single-crystal growth on transition metal substrates: how the termination and shape of hBN islands are governed by the chemical environment; how

or more than 50 years, hexagonal boron nitride (hBN) has been employed in applications exploiting its high thermal conductivity, chemical stability, refractory nature, high electrical resistivity, and lubricity.1−3 In these applications, powders, polycrystalline ceramics, and pyrolytic forms are sufficient. Grain, domain, and particle sizes can be small, and no control of the hBN’s crystallographic orientation is required. However, over the past 10 years, new applications have been envisioned, which aim to utilize hBN’s optical hyperbolicity,4 large thermal neutron capture cross-section (for the 10B isotope), atomically flat surface,5 flexoelectricity,6 and electronic properties of hBN, in a wide variety of forms: bulk crystals, atomically thin layers (nanosheets), and nanotubes. Potential applications exploiting these properties include hBN deep ultraviolet light emitters (i.e., light-emitting diodes and laser diodes),7,8 neutron detectors,9,10 nanophotonics,3,11 and singlephoton emitters.12 For these applications, control of the © 2017 American Chemical Society

Received: October 6, 2016 Accepted: March 20, 2017 Published: March 20, 2017 3585

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

www.acsnano.org

Article

ACS Nano

Figure 1. Top and side views of optimized B, N, BN, BNB, NBN, and B2N2 adsorptions on (a) Ni(111) and (b) Ni(211). Adsorption sites are indicated on the respective clean (111) and (211) surface. The N, B, Ni, and edge Ni atoms are represented by the blue, pink, gray, and green spheres, respectively.

the defect structures are induced during growth; and how the hBN nucleation and growth are influenced by the crystal facets and electronic character of the substrate. Density functional theory (DFT) calculations have matured into a powerful tool to accurately characterize the atomic structures and chemical bonding in relation to hBN synthesis. ́ et al.27 used periodic DFT For instance, Grad et al.26 and Diaz modeling to examine hBN monolayer adsorption on different transition metal substrates. For instance, on Ni(111), the nitrogen species in hBN show preference on the Ni top sites, while B prefers either the fcc or the hcp sites. DFT calculations have also been employed to identify the favorable edge termination and the shape of hBN islands.28 Furthermore, the decomposition chemistry of hBN precursors (e.g., BH2NH2) and relationship with the compositions of substrates (e.g., Ni, Ni−Cu alloy) have been modeled.13 Most recently, using firstprinciples-based analysis, Zhang et al.29 explored the shapes of hBN islands at varying boron chemical potentials on Cu(111) and Ni(111) with detailed growth kinetics of hBN edge structures. However, the computational cost inherent to the quantum mechanical approach often limits the modeling of materials synthesis on a system scale, thus rendering explicit DFT calculations inefficient or even unfeasible for understanding the

dynamic evolution leading to the hBN lattice formation. Hence, it is highly desirable to develop the capability to extend the modeling power of quantum mechanics and to ultimately address the key outstanding issues outlined previously. Reactive molecular dynamics (rMD) simulation, based on classical formulations for chemical bond formation, is an appealing alternative since it can provide dynamic perspectives of crystal growth by simulating large molecular systems at much lower computational costs compared to quantum mechanical calculations. With the combination of periodic DFT calculations and rMD simulations, the efforts to develop a selfconsistent modeling utility is presented and demonstrated. In this work, the ReaxFF reactive force field30 will be adopted and developed for the Ni/B/N system to investigate the growth mechanism of atomically thin hBN on crystalline Ni substrates. First, elementary events involved in the hBN growth, such as atomic B and N adsorptions, diffusions, and B−N bond formation producing BxNy (x, y = 1, 2) were studied in the periodic DFT setting. The governing reaction energetics and kinetics were obtained on Ni(111) and Ni(211), to account for the chemistries taking place on both terraces and step edges of the substrates. Parametrization of the ReaxFF for the B−Ni and N−Ni atomic pairs was then carried out. With the simulated hBN growth from elemental B and N deposited onto a single3586

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

ACS Nano

at their respective preferred 3-fold hcp/fcc sites (Table 1). Hence, as the molecular length of BxNy grows, the binding sites are separated further apart. As shown in Figure 1a, B2N and BN2 bind on the two next nearest 3-fold sites, while B2N2 binds on the sites separated by two 3-fold sites. The middle sections of BNB, NBN, and B2N2 are not in direct contact with the surface and buckled as illustrated in the Figure 1a site views. On Ni(211), the BxNy species are generally more strongly bound than on the Ni(111) surface, as shown in Table 1, due to the low-coordinated Ni on the Ni(211) step edge. The B and N atoms prefer the 4-fold site near the edge, as shown in Figure 1b. The BN dimer binds on Ni(211) with its N- and B-ends on the bridge sites of the edge and the lower terrace, respectively. The binding energy of BN is 1 eV stronger than that on the terrace. The linear BNB, NBN, and B2N2 molecules all prefer to bind parallel to the edge to maximize the interactions with the low-coordinated edge atoms. The terminal B and N atoms bind at the 4-fold sites for B2N and BN2, as shown in Figure 1b, respectively. For B2N2, the B-end binds at the 4-fold site, while its N-end binds at the 3-fold site of the lower terrace. The binding energies of sublayer B and N species are comparable to those in the (111) sublayer, showing less influence from the low-coordination sites on the surface. Surface Diffusions of B and N on Ni. The diffusions of hBN constituent B and N species are also critical to crystal nucleation and growth.33 Therefore, DFT calculations were performed to investigate different elemental B and N diffusion pathways to reveal their behaviors on the surface, in the sublayer, and in the bulk region of the Ni substrate. The proposed overall B and N diffusion schemes in this study are presented in Figure 2, where the B/N atom is moved as a probe

crystal Ni substrate model, the rMD simulation trajectories were then analyzed to derive the elementary growth mechanism at the atomistic level. Finally, key findings obtained from the rMD simulations were validated by additional DFT calculations.

RESULTS Adsorptions of hBN Forming Building Blocks on Ni(111) and Ni(211) Surfaces. The adsorptions of elemental B and N occur upon the release from the decomposition of Band N-containing precursors (e.g., diborane and ammonia)13,17 and thus can be considered as the first step toward hBN lattice formation. In this section, the adsorption geometries and energetics of atomic B and N and representative hBN buildingblock units were considered using periodic DFT calculations. The nickel substrates were modeled using the close-packed and stepped (111) and (211) facets to approximate the respective flat terrace and step substrate surfaces, which are ubiquitous in experimental environments.31,32 The optimized structures of adsorbed BxNy species (x = 0−2; y = 0−2) on their preferred sites on Ni(111) and Ni(211) are illustrated in Figure 1a and b, respectively. The binding energies (BEs) were calculated according to eq 1: BE = E Bx Ny − E Ni − E Bx Ny (1) * where EBxNy*, ENi, and EBxNy are the total energies of the adsorbed BxNy* species, clean Ni surface, and the total energy of corresponding BxNy species in the gas phase, respectively. These calculations, corresponding to each preferred adsorption site, are summarized in Table 1. Table 1. DFT-Calculated Binding Energies of BxNy Species (x = 0, 1, 2; y = 0, 1, 2) Adsorption on Ni(111) and Ni(211) Surfaces and Ni Sublayer Ni(111)

Ni(211)

species

site

Eads/eV

site

Eads/eV

N B N (sublayer) B (sublayer) BN

fcc hcp octahedral octahedral hcp(B)fcc(N) hcp(B)hcp(B) fcc(N)fcc(N) hcp(B)fcc(N)

−5.29 −5.84 −4.65 −6.47 −6.51

4-fold 4-fold octahedral octahedral 4-fold

−5.58 −6.95 −4.55 −6.48 −7.57

−6.11

4-fold(B)-4-fold(B)

−7.20

−7.10

4-fold(N)-4-fold(N)

−7.97

−7.25

4-fold(B)-4-fold(B)fcc(N)

−7.78

B2N BN2 B2N2

Figure 2. Schematic illustration of atomic B and N diffusion pathways on the surface, in the sublayer, and in the bulk region of a model Ni surface consisting of terrace and step sites. The solid yellow path, steps 1−4, represents surface diffusions. The dashed green path, steps 5−7, represents the sublayer diffusion. The dashed red path, steps 8 and 9, represents the bulk diffusion. The gray spheres represent the Ni atoms. The orange sphere represents the diffusing atom.

On Ni(111), B atoms prefer to bind at the 3-fold hcp site, while N atoms prefer the 3-fold fcc site. In the Ni sublayer, atomic B and N both prefer the octahedral site. However, the sublayer B atom is even more energetically stable, by 0.63 eV, compared to its surface adsorption. On the other hand, the sublayer N is less energetically favorable by 0.64 eV. The predicted atomic adsorption preference is in good agreement with a previous study by Zhang et al.29 The DFT-calculated energetics suggest that, thermodynamically, elemental B would prefer the octahedral lattice spacing in the sublayer region. Among other building-block species, the BN dimer prefers to bind on adjacent hcp(B)-fcc(N) dual sites on Ni(111). BNB, NBN, and B2N2 all bind with the terminal B and/or N species

from one 4-fold step edge site (labeled as A) to another 4-fold step edge site (labeled as B) on a model surface terminated with both terrace and step sites. The (211) step edge site was used as the starting and ending points because it is the most stable binding site for both B and N atoms (Table 1). As illustrated in Figure 2, the solid yellow path represents the diffusion pathway on the surface, which consists of four 3587

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

ACS Nano elementary diffusion steps: (1) from one 4-fold site to the neighboring 4-fold site, (2) from the 4-fold site to the 3-fold hcp/fcc on the terrace, (3) from the 3-fold site to the neighboring 3-fold fcc/hcp site on the terrace, and (4) from the fcc/hcp site to the step 4-fold site. With a similar notation, the green dashed path represents B/N diffusion into the substrate sublayer. It consists of three steps: (5) from the 3-fold site to the octahedral site in the sublayer, (6) from one octahedral site to the next octahedral site in the sublayer, and (7) from the octahedral site in the sublayer emerging at the 4-fold step edge site. The bulk diffusion pathway is represented by the red dashed path, consisting of two steps: (8) from the octahedral site in the sublayer to the octahedral site in the bulk and (9) from one bulk octahedral site to the next octahedral site in the bulk. The potential energy surfaces (PES) describing the atomic B and N diffusion energetics and kinetics are established in Figure 3a and b, respectively, using the DFT-calculated energy barriers

Table 2. DFT-Calculated Energy Barriers of Atomic B and N Diffusion Corresponding to the Elementary Steps Labeled in Figure 2 N diffusion step step step step step step step step step step

1 2 3 4 5 6 7 8 9

(step edge → step edge) (step edge → terrace) (terrace → terrace) (terrace edge → step edge) (terrace → subsurface) (subsurface → subsurface) (subsurface → step edge) (subsurface → bulk) (bulk → bulk)

B

Eaa/eV

a

Ea /eV

1.38 0.56 0.56 0.44 1.42 1.00 0.18 1.76 1.76

1.03 1.05 0.22 0.94 1.13 0.34 0.36 1.74 1.74

Energy barriers (Ea) were calculated according to Ea = ETS − EIS, where ETS and EIS represent the total energies of the transition state and the initial state, respectively. a

Therefore, unlike B diffusion, the surface diffusion pathway is clearly much more competitive. Force Field Development. ReaxFF for B−N pair interactions has been developed and applied to modeling BH3−NH3 decomposition and single-walled BN nanotube formation by Han, Weismiller, van Duin, and co-workers.34−36 Furthermore, the ReaxFF force field for single-crystal-phase nickel also exists.37−39 This work focuses on the establishment of the N−Ni and B−Ni pair potentials that will extend the capability of simulating the elementary chemical events in relation to the hBN growth process on a Ni substrate. Periodic DFT calculations, based on the Ni(111) surface, were converted into a ReaxFF training set, consisting of the adsorption structures of the BxNy building-block species, binding energies, reaction energies of the above basic building-block formations, atomic B/N diffusion barriers, and B−N bond formation energy barriers. The force field parameters were then optimized to reproduce the DFT data. The full ReaxFF training set and force field parameters are provided in the Supporting Information. In Figure 4a, the binding energies of BxNy (x, y = 0, 1, 2, 3) on Ni(111), in the sublayer, and bulk obtained from DFT calculations and ReaxFF predictions are shown. The training puts particular emphasis on the data located at the global energy minima, such as B at the sublayer octahedral site (B_sub) and N at the fcc site (N_fcc) on Ni(111). In Figure 4b, the DFT-calculated elementary step energy barriers, describing diffusion and bond formation on the Ni(111) surface and in Ni bulk for building-block species, were considered. The reaction energies corresponding to the formation of building-block species (ΔEBxNy), based on eq 2, were included.

Figure 3. PES depicting the (a) B diffusion and (b) N diffusion pathways illustrated in Figure 2. The overall diffusion energy barrier, relative to the zero potential energy reference, for each path is labeled numerically.

of the elementary steps (1−9) listed in Table 2. The overall energy barrier for B surface diffusion, referenced to the state at location A, occurring at step 4, is 1.99 eV. In comparison, the overall energy barrier of sublayer diffusion is 2.18 eV, occurring at the diffusion into the sublayer, i.e., step 5, only 0.19 eV higher than the surface diffusion pathway and exergonic. Therefore, the sublayer diffusion pathway should be competitive to atomic B surface diffusion and may be able to assist B migration. B diffusion in the bulk has the highest overall energy barrier of 2.42 eV (step 9) and is the least competitive diffusion pathway. The overall energy barrier for N diffusion on the surface is 1.38 eV, occurring at step 1, while the energy barriers for subsequent steps are rather modest. However, the overall barriers for sublayer and bulk diffusion pathways are 2.20 and 3.33 eV, corresponding to steps 5 and 9, respectively.

ΔE Bx Ny = E Bx Ny + E − EA − E B (2) * * where EA and EB refer to the total energies of reactants A and B at their most stable configurations on Ni(111), respectively. In this training, the reaction energetics of the hBN growth elementary steps was considered as well, as shown in Figure 4c. Reactive Molecular Dynamics Simulations of hBN Nucleation and Growth on the Ni(111) Surface. Figure 5 shows the evolution of the hBN growth process on a Ni(111) surface from the ReaxFF-based rMD simulation performed at 1300 K, which is comparable to experimental CVD conditions, under which high-quality hBN crystal formation has been 3588

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

ACS Nano

Figure 4. Comparisons between DFT (blue) and ReaxFF (red) for (a) binding energies of atomic B and N adsorptions; (b) energy barriers of B/N diffusions and B−N bond formations; and (c) reaction energies.

reported.40−42 Initially, the randomly deposited B and N atoms on the nickel substrate started to form BN dimers, trimers (e.g., BNB), and short chains (e.g., BNBN), which are illustrated in Figure 5a, at t = 25 ps. With the increase of the deposited B and N atoms, long linear and branched BN chains up to 10 atoms (highlighted in the red dashed circle in Figure 5b) have formed, and the central atoms (B or N) of each Y junction resemble an sp2-type hybridization. In this simulation, the first hexagon ring appears at 64.25 ps (highlighted in yellow in the red dashed circle in Figure 5c), which serves as the initial nucleus and then grows into a larger hBN island consisting of four connecting hexagons at 434.1 ps (highlighted in the red dashed circle in Figure 5d). A closer examination of the structural evolution in relation to initial hBN nucleation is shown by the isolated structures I−VI in the Figure 5 inset: a ramose “X-type” BN species in II is first formed through the coalescence of two linear BN chains formed and illustrated in I. Following structure II, a hexagon is formed (III), as a result of the folding of the two branches. This hBN hexagon, with alternating B and N atoms, has multiple dangling branches, which indeed facilitate the formation of the next hexagon by simply bending the longer branch. Subsequently, the third and fourth hexagons coalesce through a similar process, i.e., folding of two branches, as illustrated in IV and V. Hence, multiple fused hexagon clusters, consisting of two to four rings, are formed and shown in Figure 5d. Simulations performed up to 1.793 ns, as depicted in Figure 5e, showed that a continuous, atomically thin hBN network is established on the Ni surface. At 6.25 ns, the surface is mostly covered by a continuous hBN sheet dominated by perfect BN hexagons (see Figure 5f). The hBN formation process (0.25− 1.25 ns) is also shown in a video available in the Supporting Information.

The pentagon−heptagon (5|7) dislocation structure, consisting of adjacent five-membered and seven-memebered rings, as shown in Figure 6, have also been noticed in the simulation, in this case at 5.38 ns at 1300 K. This (5|7) dislocation, containing homoelemental B−B or N−N bonds, has been predicated theoretically as a main type of grain boundary in hBN crystals35,43 and also experimentally observed by Gibb et al.25 and Wang et al.44 As displayed in the dashed box with the participating atoms numerically labeled, this (5|7) structure, consisting of a homoelemental B−B bond (between the 1−2 atom pair) and an N−N bond (between the 3−4 atom pair), is located at the edge of a larger hBN fragment, highlighted by the dashed circles in Figure 6. It is possible that this (5|7) dislocation occurs at the hBN growth frontier. The analysis of the rMD trajectory revealed that the (5|7) dislocation structure is able to evolve into two joint hexagons after 2 ps. In the illustrated snapshot, a close examination of the transformation showed that the B−B bond distance is 2.208 Å, elongated from the typical B−B bond length of 1.924 Å.45 At 5.382 ns, atom 3 moves closer to atom 1, forming the energetically more favorable heteroelemental B−N bond, while also breaking the N−N bond (3−4 atom pair and 1−2 atom pair). At this time, another B−N bond is formed between atoms 2 and 4, resulting in two adjacent hexagonal geometries with alternating B−N bonds. Diffusion and Distribution of Sublayer B and N Atoms. The analyses of the atomic B and N distributions in the sublayers (second to fifth layers) of the Ni substrate have been performed, and the elemental distributions, as a function of simulation time (in ns), are shown in Figure 7a,b, respectively. At the beginning of the simulation (i.e., within the first 50 ps), equal numbers of B and N atoms are deposited, and the second layer B and N atoms experience a sharp increase 3589

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

ACS Nano

Figure 5. Snapshot images (both top and side views) taken from the trajectories of rMD simulations at 1300 K at various time frames: (a) t = 25.0 ps, (b) 47.5 ps, (c) 64.25 ps, (d) 434.0 ps, (e) 1.793 ns, and (f) 6.25 ns. The blue and pink spheres represent N and B atoms, respectively. The atoms in the Ni substrate are represented in gray for clarity. The yellow spheres in (c) illustrate the first hexagonal ring formed in this particular MD run. Both the top and side views are shown, and the B and N atoms in the interior of the Ni substrate can also be seen. The inset figures (I−VI) highlight the process of the first hexagon formation and subsequent hBN nucleation.

Figure 6. Transformation of pentagon−heptagon (5|7) dislocation structure into joint hexagons at the respective 5.38 ns and the 5.382 ns time frames during the rMD simulation at 1300 K. The local atoms involved are further highlighted with dashed circles, and the participating atoms in the rearrangements of the structures are labeled in the inset dashed box in the middle, where the B−B and the N−N bond (in the 1− 2 and 3−4 atom pairs) that are broken are marked with red crosses, and the B−N bonds (between the 1−3 and 2−4 atom pairs) formed are indicated with dashed double-headed arrows. The B and N atoms are represented by pink and blue spheres, while atoms in the Ni substrates are in gray.

(shown in blue in Figure 7a,b), as a result of the momenta of impinging atoms. After the deposition, the second layer N atoms quickly diffuse onto the surface. The number of second layer N fluctuates at around 1 after 1 ns and remains so for the rest of the simulation (Figure 7a). This is supported by our DFT calculations that show atomic N at the sublayer octahedral

site is metastable by 0.64 eV and would rather be at the 3-fold fcc site on the Ni(111) surface. In contrast, the total number of B atoms in the second layer continues to rise even after the deposition until 0.5 ns (Figure 7b), further contributed by B diffusion from the surface to the sublayer. Then, the number of B atoms fluctuates and stabilizes at around 8 for the rest of the 3590

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

ACS Nano

Figure 7. Distributions of (a) N and (b) B atoms in Ni sublayers as a function of the simulation time (in ns) at 1300 K. The blue, orange, gray, and yellow colors represent the second, third, fourth, and fifth substrate layer, respectively.

Figure 8. Snapshots taken from the trajectories of the rMD simulations at 6.25 ns depicting hBN formations (200 boron and 200 nitrogen) on Ni substrates at temperatures of (a) 900 K, (b) 1100 K, (c) 1300 K, and (d) 1500 K, respectively. The B and N atoms are represented by pink and blue spheres, while atoms in the Ni substrates are in gray.

experimental measurements performed by Kowanda and Kubota et al., which have shown that B has a much higher solubility in Ni than N.46,47 hBN Growth at Different Temperatures. Simulations were also performed at 900, 1100, and 1500 K, while the number of deposited B and N and other simulation setup remained the same. The snapshot images, taken from the final time frame of each simulation (i.e., 6.25 ns) as shown in Figure 8, illustrate the variations of the growth states at different temperatures. At 900 K (Figure 8a), most B and N species exist in small hBN islands consisting of fused hexagons, along with a

simulation. The behavior of sublayer B species can also be explained by the DFT calculations, where the B atom prefers the sublayer octahedral site and is 0.63 eV more stable than the surface hcp site. The B and N atoms in the third to fifth layers are much more scarce, reflecting less penetration during the deposition and also the much higher energy barriers of bulk diffusion. The number of N atoms in these sublayers remains almost unchanged, while the number of B atoms slightly decreases throughout the course of the simulation at 1300 K. The higher B concentration in the substrate sublayer region obtained from rMD can also be supported by a number of 3591

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

ACS Nano

nucleation and formation of a hexagonal BN structure on a Ni substrate, periodic DFT calculations were performed to confirm the linear−branch−ring mode from the reaction energetic perspective. These molecular models constructed for the DFT calculations are inspired by the motifs from previous rMD simulations and will prove to be an effective strategy to obtain insights by integrating with first-principles calculations in future studies. For quantitative analysis, a stoichiometric B−N pair was used to represent the growth unit (i.e., with an increment of BN per step). The adsorption energy (AE) was calculated according to eq 3:

number of pentagons, heptagons, and octagons. In addition, open structures with branches are also abundant. At 1100 K, hBN domains with larger size, such as the upper right corner of Figure 8b, have formed. At 1300 and 1500 K (Figure 8c,d), structures with a continuous hBN lattice, throughout the substrate surfaces, are well established. Furthermore, the numbers of hexagons throughout the simulation (up to 6.25 ns) were collected for each temperature and compared in Figure 9. As expected, the numbers in each

AE = (E Bx Nx − E Ni − xE BN)/x *

(3)

where EBxNx* and ENi are the respective total energies of the adsorbed (BN)x species and the clean Ni surface. EBN is the total energy of the BN dimer in the gas phase, and x denotes the stoichiometry. The nucleation process of hBN is considered to start from the most stable BN dimer on the Ni(111) surface, as shown on the leftmost side of Figure 10. A B2N2 molecular chain is formed with the addition of a second B−N pair. In the following steps, three types of geometrieslinear, ring, and branched configurations (Figure 10a−c)were considered; only the structures with the lowest AE in each type are shown in the figure. The linear (BN)x (x = 3−5) structures bind with their terminal B and N ends, with the atoms in the middle section buckled and detached from the substrate. The B and N terminals prefer the hcp and fcc sites, respectively, consistent with the trend noted from DFT calculations performed on the building-block species. Meanwhile, the corresponding AEs become more energetically favorable as the chain length grows. For the smallest isolated ring, (BN)3, the constituent N and B atoms bind close to the bridge sites. As the size and number of rings grow, the N atom binds at either the bridge or the neartop site, and some B atoms bind at the 3-fold site, as illustrated in (BN)4 and (BN)5 in Figure 10b. Also, the structure becomes more energetically stable as the ring grows. For (BN)5, the joint hexagonal structure is energetically favored over the single 10membered ring and is approximately 0.3 eV more stable than (BN)3 per B−N pair. Nevertheless, the calculated AEs suggest that the overall ring configurations are approximately 0.3−0.5 eV more metastable than their linear counterparts.

Figure 9. Number of hexagons as a function of the simulation time (in ns) for temperatures of 900 K (yellow), 1100 K (gray), 1300 K (orange), and 1500 K (purple), respectively.

case increase monotonically with both the simulation temperature and time. At 900 K, the number of hexagons reached 20 at 2 ns (i.e., 10 hexagons/ns based on a simple linear approximation) and then nearly plateaued. At 1100 K, the number of hexagons continues to increase up to 40 (gray) and stabilizes at approximately 3 ns (i.e., 13.3 hexagons/ns). Overall, the number of hexagons has nearly doubled compared to that at 900 K within the same time frame. At 1300 and 1500 K, there are approximately 100 (Figure 9, orange) and 110 (Figure 9, purple) hexagons, where the theoretical maximum based on the stoichiometric B and N deposited in the simulation (200 each) is 200. In addition, the two growth regimes, i.e., (1) a rapid initial growth before respective 1.8 ns (i.e., 41 hexagons/ns) and 1.5 ns (i.e., 53 hexagons/ns) and (2) slow growth afterward, at an approximate rate of 5.8 and 6.3 hexagons/ns, become more distinct compared to the profiles obtained at 900 and 1100 K. Confirmation of Nucleation and Growth Mechanism with DFT Calculations. To further validate the simulated

Figure 10. Optimized geometries and corresponding adsorption energies (shown underneath each configuration) for (a) linear, (b) ring, and (c) branched configurations from BN to B5N5 on Ni(111). The B, N, and Ni atoms are represented by pink, blue, and gray spheres, respectively. 3592

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

ACS Nano

CONCLUSIONS In order to elucidate the mechanism of hBN growth on Ni, DFT calculations were first performed to obtain the energetics and kinetics of BN species on Ni(111) and Ni(211) surfaces. We showed that B atoms prefer the sublayer octahedral site, while N atoms prefer the 3-fold fcc surface sites. In addition, the adsorbed BxNy (x = 0−2; y = 0−2) species bind more strongly to the low-coordinated Ni in the Ni(211) step edge. DFT calculations have also indicated that surface diffusions are still kinetically favored for both B and N atoms. However, sublayer diffusion can also be competitive for B atoms. These DFT calculation data were subsequently used to construct a quantum mechanical training set for Ni−B and Ni− N pair potentials. Furthermore, rMD simulations of B/N deposition on Ni(111) were performed at temperatures ranging from 900 to 1500 K. Detailed examination of rMD trajectories showed that linear and branched BN species are formed first on the Ni surface. Subsequently, the first BN hexagon is formed as the branches fold, acting as the nucleus for further hBN growth. The restructuring of (5|7) dislocation can take place at 1300K through N−N and B−B bond breaking, accompanied by B−N bond formations. The atomic B and N distribution analysis is consistent with DFT calculations, indicating that B is more abundant in the interior of the Ni substrate than N. By comparing the number of hexagons at different temperatures, our simulations have demonstrated the formation of continuous, atomically thin hBN on the Ni substrate can be favored at higher temperature (e.g., 1300 and 1500 K) within the simulation time frame considered in this study. Additional DFT calculations have been performed to validate the nucleation process observed during the rMD simulation, with the outcome supporting the mechanism derived from the newly developed ReaxFF potential.

For the branched structures, the additional B−N pairs were added to the middle B2N2 or (BN)3. As shown in Figure 10c, the branched configurations also tend to bind with their terminal B and N atom species at their preferred 3-fold sites. Like the linear structure, there is no direct contact between the atoms in the middle section of the structure with the Ni substrate. The AE also decreases as (BN)x grows. For x = 3−5, the branched configurations are not as energetically favorable as the linear configurations with the same number of BN units, but slightly more favorable than the ring configurations. The above analysis indicates that, for BN unit numbers, the linear structures dominate the initial nucleation stage of hBN growth, consistent with the structures observed from the rMD simulations. Similarly, such behavior has been predicted for early stage graphene formation from its elemental constituent carbon species by Li et al.,48 who claim that the branches formed from the linear carbon chains. Given the energetic preference for linear BN chains at the initial nucleation stage, two possible ways are proposed to form hexagonal rings: (1) the formation of a branch-type structure through the combination of linear chains, which then fold, and (2) the bending of a single linear chain. As shown in Figure 11,

METHODS Figure 11. Energy profile showing the variations of adsorption energies as a function of the number of BN pairs in the BN units considered. The B and N atoms are in pink and blue, respectively.

Density Functional Theory Calculations. All periodic DFT calculations were performed with spin polarization using the Vienna Ab initio Simulation Package (VASP) program.49 The pseudopotentials for B, N, and Ni elements were generated from the projector augmented wave (PAW)50 method, with the plane-wave basis set energy cutoff at 400 eV. The generalized gradient approximation (GGA) Perdew−Burke−Ernzerhof (PBE)51 functional was employed to account for the electron exchange−correlation potential. The Brillouin zone integrations were carried out using a 4 × 4 × 1 Monkhorst−Pack mesh.52 A Methfessel−Paxton smearing53 of 0.2 eV was used, with the total energies then extrapolated to 0 K. The geometry optimization stops when all the forces on the system become smaller than 0.02 eV/Å. The bulk Ni lattice constant and magnetic moment from the bulk optimization are 3.52 Å and 0.62 μB/Ni atom, respectively, in good agreement with experimental values of 3.52 Å and 0.61 μB.54 The Ni surface was modeled by considering the close-packed terrace sites and the stepped sites, where the single-crystal Ni(111) and Ni(211) facets were used. The Ni(111) and Ni(211) surfaces were represented by a four-layer p(3 × 3) unit cell and a 12-layer Ni(211) slab with a p(1 × 3) unit cell, respectively, for periodic DFT calculations. For respective (111) and (211) slabs, the bottom two layers and the bottom six layers were fixed at their optimized bulk lattice values. A 30 Å thick vacuum space was used for both unit cells to avoid interference between periodic images along the perpendicular direction to the surface. The climbing-image Nudged Elastic Band (CINEB)55 and the Dimer56 calculations were performed in order to identify the minimum energy pathways and the transition-state structures of the elementary nucleation and growth steps. Each

the AEs of branch-type BN species (x = 3−5) are closer to the linear chains compared with the ring type; thus it is likely that the next step following the formation of linear chians is to form branched structures. This is consistent with the observed hBN nucleation process noted in rMD simulations. Once the branched BN speices are formed, the formation of the first hexagon (x = 6) becomes energetically favored. As x becomes sufficiently large, i.e., x = 9, the AE of the structure with joint hexagonal rings becomes more and more favorable. The lower AE limit (i.e., −8.81 eV/B−N pair), in Figure 11, corresponds to the periodic hBN on the Ni(111) surface and reveals that the overall energetics will eventually drive the process to form the hBN monolayer. Compared to graphene growth, the heteratomic structure of hBN has definitely introduced additional chemical complexity in the nucleation and growth mechanism. In this study, the trend predicted by deterministic rMD simulations has shown rather satisfactory agreement with detailed DFT analysis and provided clear evidence that the linear−branch−ring pathway can indeed be used to describe hBN formation from elemental B and N. 3593

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

ACS Nano transition state was confirmed by the vibrational frequency analysis with a single imaginary frequency. ReaxFF Force Field. ReaxFF is an empirical pairwise potential, in which the energy contributions from chemical bonding, valence angle, and torsion angles are formulated as functionals of the bond order, which is expressed as a function of atomic-pair distance. The detailed mathematical expressions for ReaxFF were presented in ref 30. As a general form, the total system energy in ReaxFF can be partitioned into the expression below, as shown eq 4:

Esys = E(valence) + E(Coulomb) + E(vdW)

ACKNOWLEDGMENTS Support from the Materials Engineering and Processing program of the National Science Foundation, award number 1538127, and the II−VI Foundation is greatly appreciated. DFT and rMD calculations were carried out thanks to the supercomputing resources and services from the Center for Nanoscale Materials (CNM) supported by the Office of Science of the U.S. Department of Energy under contract no. DE-AC02-06CH11357; the Beocat Research Cluster at Kansas State University, which is funded in part by NSF grant CNS1006860; and the National Energy Research Scientific Computing Center (NERSC) under contract no. DEAC0205CH11231.

(4)

where E(valence), E(Coulomb) , and E(vdW) represent the valence, Coulombic, and van der Waals contributions to the system energy, respectively. E(valence) itself is a sum of all the bond order dependent terms, including the bonding (e.g., Ni−B, Ni−N), valence angles (e.g., Ni−N−Ni, Ni−B−Ni, B−N−Ni), and torsion angles (e.g., Ni−B−N− Ni), in the ReaxFF formulation. This term will disappear as the associated chemical bond is broken (i.e., bond order becomes zero). The parameters in eq 4 are determined using quantum mechanical data. The bond, angle, and torsion parameters for B−B, N−N, and B− N pairs and bulk Ni systems were adopted from previous work.35 The complete set of the ReaxFF force field is included in the Supporting Information. Reactive Molecular Dynamics Simulation Setup. The rMD simulations of hBN growth on the Ni surface were performed with LAMMPS.57 A rectangular-shaped nickel slab with the close-packed (111) facet was used to represent the Ni substrate, where elemental B and N were deposited initially for hBN nucleation and growth. The Ni slab consists of five atomic layers, with lateral dimensions of 12 × 12 Ni atoms, for a total of 720 Ni atoms. Periodic boundary conditions were imposed on all dimensions. The bottom (fifth) layer was fixed so that the Ni atoms remained in their bulk lattice structures during the simulation to represent the interior bulk region, while the top four layers were relaxed along the vertical direction to represent the surface and sublayer regions. There is a vacuum of 90 Å above and below the slab. In the current simulation procedure, an equal number of B and N atoms (200 each), based on the stoichiometric ratio of hBN, were sequentially introduced in pairs from the gas phase with random x, y coordinates above the Ni surface at an interval of 0.25 ps. To minimize premature B−N bond formation, the distance between the initial B and N sources is set to be at least 1.90 Å, which is much larger than the typical BN bond length of 1.44 Å. All the B and N were deposited only on the relaxed side of the slab with controlled initial momenta. The simulations were run at different temperatures, i.e., 900, 1100, 1300, and 1500 K, controlled by a Nose Hoover thermostat.58 Each MD simulation was run for at least 6 ns using a time step of 0.25 fs. The equation of motion was numerically solved using the velocity Verlet integration scheme.59

REFERENCES (1) Raman, C.; Meneghetti, P. Boron Nitride Finds New Applications in Thermoplastic Compounds. Plast. Addit. Compd. 2008, 10, 26−31. (2) Lipp, A.; Schwetz, K. A.; Hunold, K. Hexagonal Boron Nitride: Fabrication, Properties and Applications. J. Eur. Ceram. Soc. 1989, 5, 3−9. (3) Taylor, K. Hot pressed boron nitride. Ind. Eng. Chem. 1955, 47, 2506−2509. (4) Dai, S.; Ma, Q.; Andersen, T.; Mcleod, A.; Fei, Z.; Liu, M.; Wagner, M.; Watanabe, K.; Taniguchi, T.; Thiemens, M. Subdiffractional Focusing and Guiding of Polaritonic Rays in a Natural Hyperbolic Material. Nat. Commun. 2015, 6, 6963. (5) Dean, C. R.; Young, A. F.; Meric, I.; Lee, C.; Wang, L.; Sorgenfrei, S.; Watanabe, K.; Taniguchi, T.; Kim, P.; Shepard, K. Boron Nitride Substrates for High-Quality Graphene Electronics. Nat. Nanotechnol. 2010, 5, 722−726. (6) Naumov, I.; Bratkovsky, A. M.; Ranjan, V. Unusual Flexoelectric Effect in Two-Dimensional Noncentrosymmetric sp2-Bonded Crystals. Phys. Rev. Lett. 2009, 102, 217601. (7) Watanabe, K.; Taniguchi, T.; Niiyama, T.; Miya, K.; Taniguchi, M. Far-Ultraviolet Plane-Emission Handheld Device Based on Hexagonal Boron Nitride. Nat. Photonics 2009, 3, 591−594. (8) Jiang, H.; Lin, J. Hexagonal Boron Nitride for Deep Ultraviolet Photonic Devices. Semicond. Sci. Technol. 2014, 29, 084003. (9) Ahmed, K.; Dahal, R.; Weltz, A.; Lu, J.; Danon, Y.; Bhat, I. Growth of Hexagonal Boron Nitride on (111) Si for Deep UV Photonics and Thermal Neutron Detection. Appl. Phys. Lett. 2016, 109, 113501. (10) Maity, A.; Doan, T.; Li, J.; Lin, J.; Jiang, H. Realization of Highly Efficient Hexagonal Boron Nitride Neutron Detectors. Appl. Phys. Lett. 2016, 109, 072101. (11) Caldwell, J. D.; Kretinin, A. V.; Chen, Y.; Giannini, V.; Fogler, M. M.; Francescato, Y.; Ellis, C. T.; Tischler, J. G.; Woods, C. R.; Giles, A. J. Sub-Diffractional Volume-Confined Polaritons in the Natural Hyperbolic Material Hexagonal Boron Nitride. Nat. Commun. 2014, 5, 5221. (12) Tran, T. T.; Zachreson, C.; Berhane, A. M.; Bray, K.; Sandstrom, R. G.; Li, L. H.; Taniguchi, T.; Watanabe, K.; Aharonovich, I.; Toth, M. Quantum Emission from Defects in Single-Crystalline Hexagonal Boron Nitride. Phys. Rev. Appl. 2016, 5, 034005. (13) Lu, G.; Wu, T.; Yuan, Q.; Wang, H.; Wang, H.; Ding, F.; Xie, X.; Jiang, M. Synthesis of Large Single-Crystal Hexagonal Boron Nitride Grains on Cu−Ni Alloy. Nat. Commun. 2015, 6, 6160. (14) Kobayashi, Y.; Nakamura, T.; Akasaka, T.; Makimoto, T.; Matsumoto, N. Hexagonal Boron Nitride on Ni (111) Substrate Grown by Flow-Rate Modulation Epitaxy. J. Cryst. Growth 2007, 298, 325−327. (15) Chatterjee, S.; Luo, Z.; Acerce, M.; Yates, D. M.; Johnson, A. C.; Sneddon, L. G. Chemical Vapor Deposition of Boron Nitride Nanosheets on Metallic Substrates via Decaborane/Ammonia Reactions. Chem. Mater. 2011, 23, 4414−4416.

ASSOCIATED CONTENT S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsnano.6b06736. Additional information (PDF) Movie (AVI)

AUTHOR INFORMATION Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. ORCID

Bin Liu: 0000-0001-7890-7612 Notes

The authors declare no competing financial interest. 3594

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

ACS Nano (16) Auwärter, W.; Suter, H. U.; Sachdev, H.; Greber, T. Synthesis of One Monolayer of Hexagonal Boron Nitride on Ni (111) from BTrichloroborazine (ClBNH)3. Chem. Mater. 2004, 16, 343−345. (17) Ismach, A.; Chou, H.; Ferrer, D. A.; Wu, Y.; McDonnell, S.; Floresca, H. C.; Covacevich, A.; Pope, C.; Piner, R.; Kim, M. J. Toward the Controlled Synthesis of Hexagonal Boron Nitride Films. ACS Nano 2012, 6, 6378−6385. (18) Lee, Y.; Liu, K.; Lu, A.; Wu, C.; Lin, C.; Zhang, W.; Su, C.; Hsu, C.; Lin, T.; Wei, K. Growth Selectivity of Hexagonal-Boron Nitride Layers on Ni with Various Crystal Orientations. RSC Adv. 2011, 2, 111−115. (19) Kim, K. K.; Hsu, A.; Jia, X.; Kim, S. M.; Shi, Y.; Hofmann, M.; Nezich, D.; Rodriguez-Nieva, J. F.; Dresselhaus, M.; Palacios, T. Synthesis of Monolayer Hexagonal Boron Nitride on Cu Foil Using Chemical Vapor Deposition. Nano Lett. 2011, 12, 161−166. (20) Song, L.; Ci, L.; Lu, H.; Sorokin, P. B.; Jin, C.; Ni, J.; Kvashnin, A. G.; Kvashnin, D. G.; Lou, J.; Yakobson, B. I. Large Scale Growth and Characterization of Atomic Hexagonal Boron Nitride Layers. Nano Lett. 2010, 10, 3209−3215. (21) Gibb, A.; Alem, N.; Zettl, A. Low Pressure Chemical Vapor Deposition Synthesis of Hexagonal Boron Nitride on Polycrystalline Metal Foils. Phys. Status Solidi B 2013, 250, 2727−2731. (22) Guo, N.; Wei, J.; Fan, L.; Jia, Y.; Liang, D.; Zhu, H.; Wang, K.; Wu, D. Controllable Growth of Triangular Hexagonal Boron Nitride Domains on Copper Foils by an Improved Low-Pressure Chemical Vapor Deposition Method. Nanotechnology 2012, 23, 415605. (23) Nagashima, A.; Tejima, N.; Gamou, Y.; Kawai, T.; Oshima, C. Electronic Structure of Monolayer Hexagonal Boron Nitride Physisorbed on Metal Surfaces. Phys. Rev. Lett. 1995, 75, 3918. (24) Kim, G.; Jang, A.; Jeong, H. Y.; Lee, Z.; Kang, D. J.; Shin, H. S. Growth of High-Crystalline, Single-Layer Hexagonal Boron Nitride on Recyclable Platinum Foil. Nano Lett. 2013, 13, 1834−1839. (25) Gibb, A. L.; Alem, N.; Chen, J.; Erickson, K. J.; Ciston, J.; Gautam, A.; Linck, M.; Zettl, A. Atomic Resolution Imaging of Grain Boundary Defects in Monolayer Chemical Vapor Deposition-Grown Hexagonal Boron Nitride. J. Am. Chem. Soc. 2013, 135, 6758−6761. (26) Grad, G.; Blaha, P.; Schwarz, K.; Auwärter, W.; Greber, T. Density Functional Theory Investigation of the Geometric and Spintronic Structure of h-BN/Ni (111) in view of Photoemission and STM Experiments. Phys. Rev. B: Condens. Matter Mater. Phys. 2003, 68, 085404. (27) Díaz, J. G.; Ding, Y.; Koitz, R.; Seitsonen, A. P.; Iannuzzi, M.; Hutter, J. Hexagonal Boron Nitride on Transition Metal Surfaces. Theor. Chem. Acc. 2013, 132, 1−17. (28) Liu, Y.; Bhowmick, S.; Yakobson, B. I. BN White Graphene with “Colorful” Edges: The Energies and Morphology. Nano Lett. 2011, 11, 3113−3116. (29) Zhang, Z.; Liu, Y.; Yang, Y.; Yakobson, B. I. Growth Mechanism and Morphology of Hexagonal Boron Nitride. Nano Lett. 2016, 16, 1398. (30) Van Duin, A. C. T.; Dasgupta, S.; Lorant, F.; Goddard, W. A. ReaxFF: A Reactive Force Field for Hydrocarbons. J. Phys. Chem. A 2001, 105, 9396−9409. (31) Auwärter, W.; Kreutz, T.; Greber, T.; Osterwalder, J. XPD and STM Investigation of Hexagonal Boron Nitride on Ni (111). Surf. Sci. 1999, 429, 229−236. (32) Auwärter, W.; Muntwiler, M.; Osterwalder, J.; Greber, T. Defect Lines and Two-Domain Structure of Hexagonal Boron Nitride Films on Ni (111). Surf. Sci. 2003, 545, L735−L740. (33) Venables, J.; Spiller, G.; Hanbucken, M. Nucleation and Growth of Thin Films. Rep. Prog. Phys. 1984, 47, 399. (34) Weismiller, M. R.; Duin, A. C. T.; Lee, J.; Yetter, R. A. ReaxFF Reactive Force Field Development and Applications for Molecular Dynamics Simulations of Ammonia Borane Dehydrogenation and Combustion. J. Phys. Chem. A 2010, 114, 5485−5492. (35) Han, S. S.; Kang, J. K.; Lee, H. M.; van Duin, A. C. T.; Goddard, W. A., III The Theoretical Study on Interaction of Hydrogen with Single-Walled Boron Nitride Nanotubes. I. The Reactive Force Field Reaxff (HBN) Development. J. Chem. Phys. 2005, 123, 114703.

(36) Han, S. S.; Kang, J. K.; Lee, H. M.; van Duin, A. C. T.; Goddard, W. A., III Theoretical Study on Interaction of Hydrogen with SingleWalled Boron Nitride Nanotubes. II. Collision, Storage, and Adsorption. J. Chem. Phys. 2005, 123, 114704. (37) Mueller, J. E.; van Duin, A. C. T.; Goddard, W. A., III Development and Validation of Reaxff Reactive Force Field for Hydrocarbon Chemistry Catalyzed by Nickel. J. Phys. Chem. C 2010, 114, 4939−4949. (38) Liu, B.; Lusk, M. T.; Ely, J. F. Reactive Molecular Dynamic Simulations of Hydrocarbon Dissociations on Ni (111) Surfaces. Surf. Sci. 2012, 606, 615−623. (39) Liu, B.; Lusk, M. T.; Ely, J. F. Hydrogen Dissociations on Small Nickel Clusters. Mol. Simul. 2009, 35, 928−935. (40) Shi, Y.; Hamsen, C.; Jia, X.; Kim, K. K.; Reina, A.; Hofmann, M.; Hsu, A. L.; Zhang, K.; Li, H.; Juang, Z. Synthesis of Few-Layer Hexagonal Boron Nitride Thin Film by Chemical Vapor Deposition. Nano Lett. 2010, 10, 4134−4139. (41) Tsuda, O.; Watanabe, K.; Taniguchi, T. Band-Edge Luminescence at Room Temperature of Boron Nitride Synthesized by Thermal Chemical Vapor Phase Deposition. Jpn. J. Appl. Phys. 2007, 46, L287. (42) Huang, J.; Pan, C.; Lii, D. Investigation of the BN Films Prepared by Low Pressure Chemical Vapor Deposition. Surf. Coat. Technol. 1999, 122, 166−175. (43) Liu, Y.; Zou, X.; Yakobson, B. I. Dislocations and Grain Boundaries in Two-Dimensional Boron Nitride. ACS Nano 2012, 6, 7053−7058. (44) Wang, J.; Li, S.; Liu, J. Migrations of Pentagon−Heptagon Defects in Hexagonal Boron Nitride Monolayer: The First-Principles Study. J. Phys. Chem. A 2015, 119, 3621−3627. (45) Shoji, Y.; Matsuo, T.; Hashizume, D.; Gutmann, M. J.; Fueno, H.; Tanaka, K.; Tamao, K. Boron−Boron σ-Bond Formation by TwoElectron Reduction of a H-Bridged Dimer of Monoborane. J. Am. Chem. Soc. 2011, 133, 11058−11061. (46) Kowanda, C.; Speidel, M. Solubility of Nitrogen in Liquid Nickel and Binary Ni−Xi alloys (Xi= Cr, Mo, W, Mn, Fe, Co) under Elevated Pressure. Scr. Mater. 2003, 48, 1073−1078. (47) Kubota, Y.; Watanabe, K.; Tsuda, O.; Taniguchi, T. Deep Ultraviolet Light-Emitting Hexagonal Boron Nitride Synthesized at Atmospheric Pressure. Science 2007, 317, 932−934. (48) Li, Y.; Li, M.; Wang, T.; Bai, F.; Yu, Y. DFT Study on the Atomic-Scale Nucleation Path of Graphene Growth on the Cu (111) Surface. Phys. Chem. Chem. Phys. 2014, 16, 5213−5220. (49) Kresse, G.; Furthmüller, J. Efficient Iterative Schemes for ab initio Total-Energy Calculations Using a Plane-Wave Basis Set. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54, 11169. (50) Blöchl, P. E. Projector Augmented-Wave Method. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 50, 17953. (51) Ernzerhof, M.; Scuseria, G. E. Assessment of the Perdew− Burke−Ernzerhof Exchange-Correlation Functional. J. Chem. Phys. 1999, 110, 5029−5036. (52) Monkhorst, H. J.; Pack, J. D. Special Points for Brillouin-Zone Integrations. Phys. Rev. B 1976, 13, 5188. (53) Methfessel, M.; Paxton, A. High-Precision Sampling for Brillouin-Zone Integration in Metals. Phys. Rev. B: Condens. Matter Mater. Phys. 1989, 40, 3616. (54) Bhatia, B.; Sholl, D. S. Chemisorption and Diffusion of Hydrogen on Surface and Subsurface Sites of Flat and Stepped Nickel Surfaces. J. Chem. Phys. 2005, 122, 204707. (55) Henkelman, G.; Uberuaga, B. P.; Jónsson, H. A Climbing Image Nudged Elastic Band Method for Finding Saddle Points and Minimum Energy Paths. J. Chem. Phys. 2000, 113, 9901−9904. (56) Henkelman, G.; Jónsson, H. A Dimer Method for Finding Saddle Points on High Dimensional Potential Surfaces Using Only First Derivatives. J. Chem. Phys. 1999, 111, 7010−7022. (57) Plimpton, S.; Crozier, P.; Thompson, A. LAMMPS-Large-Scale Atomic/Molecular Massively Parallel Simulator. Sandia Natl. Lab. 2007, 18. 3595

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596

Article

ACS Nano (58) Evans, D. J.; Holian, B. L. The Nose−Hoover Thermostat. J. Chem. Phys. 1985, 83, 4069−4074. (59) Swope, W. C.; Andersen, H. C.; Berens, P. H.; Wilson, K. R. A Computer Simulation Method for the Calculation of Equilibrium Constants for the Formation of Physical Clusters of Molecules: Application to Small Water Clusters. J. Chem. Phys. 1982, 76, 637− 649.

3596

DOI: 10.1021/acsnano.6b06736 ACS Nano 2017, 11, 3585−3596